An authentication and access control method based on the automatic and immediate verification of a person's bodily traits is known as biometric security. Recently, Biometric Identification Systems (BIS) demands have risen significantly. However, it is challenging to offer robust security quickly because most conventional single and multimodal biometric authentication systems utilize data size, static parameters, and keys. As a result, it has various issues, including noise in the data, unacceptable error rates, and spoof attacks. To overcome these issues, this paper designs a separately extracted feature fusion based convolutional neural network with bat optimization (SEFF-CNN-BO). Additionally, the Attribute-Based Encryption (ABE) technique is employed to securely share the generated cryptographic key. Moreover, Bat Optimization and user input string-based permutation followed by SHA-256 hash value generation is applied to the extracted biometric features in the key generation phase to ensure key revocability and to enhance the security of the biometric system. This SEFF-CNN-BO system has combined biological features from the iris, face, and fingerprint for individual identification. The performance of SEFF-CNN-BO was calculated and evaluated in terms of precision, accuracy, recall, specificity, sensitivity, F-Score, etc. Compared to previous models, the developed SEFF-CNN-BO model attained 99.56% accuracy and 99.64% recall respectively. A comprehensive analysis of the proposed model against different security attacks, key revocability, and evaluation of the strength of the generated cryptographic keys is done using various metrics available in NIST statistical test suite. Analysis shows that the method is secure against all known attacks and also the key is 100% revocable.